Re: Pascal Calling Convention

[Adam Beneschan <adam@irvine.com>]

On Mar 29, 11:30 am, Robert A Duff <bobd...@shell01.TheWorld.com>
wrote:
> Hyman Rosen <hyro...@mail.com> writes:
> > And it's a good program to keep at hand to serve as a counterexample
> > when programming languages want access to uninitialized data to be
> > undefined behavior. It's more sensible to regard such data as having
> > arbitrary but fixed values, subject to a validity check.
>
> Just because you know of one algorithm that could use that semantics?!
> (OK, I know of one other case.  Sort of.)
>
> If I wanted to add this "feature" to Ada, I'd do so via an attribute:
> T'Arbitrary_Value returns an arbitrary value of type T.  Then:
>
>     My_Array := (1..1_000_000 => Integer'Arbitrary_Value);
>
> can be implemented in zero machine instructions.  You really don't want
> to use arbitrary values by accident.

If that feature were added, I'd want the language to specify that the
value of the attribute must be a valid value for the subtype.  (Maybe
that's what you meant and didn't say so explicitly?)  That would still
be a no-op if every bit pattern would be a valid value, which would be
the case for Integer (or for the first subtype of a "mod 2**32" or
"mod 2**64" type on most machines, which I think would be needed to
make Hyman's example work).

                                -- Adam